The wisdom of crowds is the idea that the combined judgment of many independent people is often more accurate than the judgment of any single expert. In a market, that combining happens through prices: every trade nudges the price, so the final number reflects what many people believe, weighted by how much they are willing to commit. It works surprisingly well when people think independently and have different information, and it breaks down when they copy each other, when a market is thin, or when everyone shares the same blind spot.

Key takeaways

  • Averaging many independent guesses tends to cancel out individual errors, which is why a crowd can beat most experts.
  • James Surowiecki named four conditions for a wise crowd: diversity of opinion, independence, decentralization and aggregation.
  • In markets, aggregation happens through price discovery: trades turn scattered private knowledge into one public number.
  • Evidence from election markets and forecasting tournaments suggests crowds are often well calibrated, though the research is still ongoing.
  • Crowds fail when people herd, when markets are thin or manipulated, when everyone shares a bias, or when the question itself is unclear.

What Is the Wisdom of Crowds?

The wisdom of crowds describes a statistical effect: when many people estimate something on their own, their individual mistakes point in different directions and partly cancel out. The average or median of the group then lands closer to the truth than most individual guesses.

This matters for prediction markets because they are built on exactly this bet. A market does not ask one analyst what will happen; it lets many people with different views trade, and treats the resulting price as a collective estimate. Researchers sometimes call these tools information markets, and they are one practical form of what is broadly known as collective intelligence. If you are new to the mechanics, our pillar guide on how prediction markets work covers the basics.

An old intuition sits behind this, too. The Condorcet jury theorem, from the 18th century, shows that if each voter is somewhat more likely to be right than wrong, and they vote independently, a larger group becomes more likely to reach the right answer. The key words are "independently" and "somewhat more likely to be right". Both will come back when we look at where crowds fail.

Galton's Ox and Surowiecki's Four Conditions

The most famous story about crowd wisdom comes from a country fair in Plymouth, England, in 1906. About 800 visitors paid to guess the weight of an ox. The statistician Francis Galton collected the tickets afterward, expecting to show how poorly ordinary people judge such things. Instead, the median guess was 1,207 pounds, and the ox actually weighed 1,198 pounds. The crowd was off by less than 1%. Galton published the result in the journal Nature in 1907.

Nearly a century later, the journalist James Surowiecki turned this and many similar examples into a book. "The Wisdom of Crowds", published in 2004, argued that groups can be remarkably smart, but only under the right conditions. He described four:

  1. Diversity of opinion. Each person brings some private information, even if it is just an unusual reading of facts everyone knows.
  2. Independence. People form their views without simply adopting the opinions of those around them.
  3. Decentralization. People can specialize and draw on local knowledge that no central planner has.
  4. Aggregation. Some mechanism turns many private judgments into one collective answer.

Surowiecki listed prediction markets among the examples of that last ingredient: a way to aggregate. The first three conditions describe the people; the fourth describes the machinery. Remove any one of them and the wisdom of the crowd tends to weaken.

How Prices Aggregate Information: Price Discovery in Plain English

So what is price discovery? It is the process by which buyers and sellers, each acting on their own information, push a price toward the level that balances their views. Nobody sets the price; it emerges from trading.

The economist Friedrich Hayek made the classic argument for why this matters. In "The Use of Knowledge in Society", published in the American Economic Review in 1945, he pointed out that useful knowledge is scattered among millions of people, and no single authority could ever gather it. Prices solve that problem without anyone coordinating. His illustration: when a raw material becomes scarcer, its price rises, and thousands of people who could never be identified start using it more sparingly. They do not need to know why it became scarce. The price tells them enough.

A prediction market applies the same logic to a question about the future. A contract pays $1 if an event happens and $0 if it does not. Someone who thinks the event is more likely than the current price implies has a reason to buy; someone who thinks it is less likely has a reason to sell. Each trade carries a small piece of private information into the price.

HYPOTHETICAL example (made-up numbers). Imagine a contract trading at 50 cents and three traders. Ana has read a detailed local report and thinks the chance is 70%. Ben follows national coverage and thinks it is 55%. Cara has seen news that worries her and thinks it is 35%. Ana buys, pushing the price up. Ben buys a little until the price passes 55 cents, then stops. Cara sells as the price rises. If the trading settles near 58 cents, that number blends all three views, weighted by how much each was willing to commit. None of them had to agree with the others.

That final price is commonly read as a probability: 58 cents suggests roughly a 58% chance. But it is an estimate, not a guarantee. It depends on who is trading, how much money is behind each view, fees and market liquidity. Our guide on how to read odds and prices explains this in more detail, and the glossary entry on implied probability gives the short version.

There is also a link to finance. The idea of market efficiency, central to the efficient market hypothesis, says prices quickly reflect available information. Prediction markets are, in a sense, a test of that idea applied to events rather than company shares.

Does It Actually Work? Evidence From Markets and Forecasting Tournaments

Stories are persuasive, but the better question is what the data shows. Several lines of evidence point in the same direction, with caveats.

Iowa Electronic Markets. Running since 1988, this small academic market lets people trade on elections with real money. A study published in 2008 compared its prices with 964 polls across five US presidential elections from 1988 to 2004. The market was closer to the final result than the polls about 74% of the time. You can read more about this period in our article on the history of prediction markets.

Wall Street betting markets. Long before scientific polling, organized election betting took place in New York. Economists Paul Rhode and Koleman Strumpf found that between 1884 and 1940, the candidate favored in mid-October went on to win 11 of 15 presidential elections.

The academic review. In a 2004 survey in the Journal of Economic Perspectives, Justin Wolfers and Eric Zitzewitz concluded that market-generated forecasts are typically fairly accurate and outperform most moderately sophisticated benchmarks. That is a careful claim: good, not perfect.

Forecasting tournaments. Markets are not the only way to aggregate opinions. In 2011, the US intelligence research agency IARPA launched a forecasting tournament on geopolitical events. One team, the Good Judgment Project, was led by Philip Tetlock and Barbara Mellers at the University of Pennsylvania. Over four years it covered about 500 questions and more than a million forecasts. According to Good Judgment, its team won the tournament and even outperformed intelligence analysts who had access to classified data. The project also identified "superforecasters": people who are consistently good at assigning realistic probabilities, even outside their own field. Their success suggests that collective wisdom improves when you combine many views and also give more weight to people with a strong track record.

The economist Robin Hanson has pushed the idea further with futarchy, a proposal to use markets to help guide policy decisions. It remains a debated idea rather than a tested system.

Where the Crowd Breaks: The Limits of Collective Wisdom

All of this evidence comes with a condition attached: the crowd has to meet those four conditions, at least roughly. When it does not, the same mechanism that produces wisdom can produce confident mistakes. Here are the main ways it goes wrong.

Herd Behavior and the "Madness of Crowds": When the Crowd Stops Being Independent

Herd behavior happens when people stop relying on their own information and start following what others do. An information cascade is an extreme version: each person reasons that the people before them probably knew something, so they copy them, and soon a whole group moves in one direction based on very little real evidence.

A 2011 experiment by Jan Lorenz and colleagues, published in PNAS, tested this directly. 144 participants answered factual estimation questions. When people could see what others guessed, even mild social influence undermined the crowd effect. Estimates became less diverse, but the group's collective error did not improve. The true answer drifted toward the edge of the range of guesses, and people felt more confident after converging, even though they were not more accurate.

The phrase "madness of crowds" comes from Charles Mackay's 1841 book "Extraordinary Popular Delusions and the Madness of Crowds", which popularized the story of tulip mania in the Netherlands, when some bulb prices rose sharply from December 1636 and collapsed in 1637. Historian Anne Goldgar's archival research suggests the real episode was much smaller than the legend: about 350 traders, no bankruptcy case she could find, and little effect on the wider economy. Mackay relied partly on satirical pamphlets. The lesson cuts both ways: crowds can get carried away, and stories about crowds can get carried away too.

Market Manipulation: Can Someone Buy the Price?

In prediction markets, market manipulation means placing trades to move the price for reasons other than a genuine view on the outcome, for example to make a candidate look stronger. It can happen, but the evidence suggests it is expensive and usually short-lived in active markets.

In a 2008 working paper, Rhode and Strumpf described a deliberate experiment in the Iowa Electronic Markets during the 2000 presidential race. They made 11 planned trades totaling about $3,116, roughly 2% of market volume. Prices rose about 4% in the first half hour, half of the effect was gone within about two and a half hours, and it had disappeared within twelve hours. The same paper counted 46 documented manipulation accusations in New York betting markets between 1880 and 1944, where prices typically bounced back within days. On the exchange TradeSports in 2004, two sudden price drops were reversed within minutes. The authors noted that recovery times shrank from days to hours to minutes as markets became faster.

A more expensive case came in 2012. Researchers David Rothschild and Rajiv Sethi, as reported by Slate in 2013, found that one trader on Intrade accounted for about one third of all bets on Mitt Romney in the final two weeks of the campaign and lost between $4 million and $7 million. They concluded the trader was trying to distort prices. The takeaway: manipulation is possible, it can be costly for the person attempting it, and because prices are visible, they can still shape media narratives while the distortion lasts.

Thin Markets and Liquidity

A crowd needs to be a crowd. When only a handful of people trade a contract, a single large order can move the price a long way, and that move may say more about one person's wallet than about the event. This is why trading volume and liquidity matter when you read a price. Deep markets usually have market makers and many participants willing to trade at close prices; thin ones may not.

Shared Biases and the Favorite-Longshot Bias

Averaging cancels out errors only when those errors point in different directions. If the whole crowd shares the same blind spot, averaging just preserves it. One well-known pattern comes from betting markets: the favorite-longshot bias, first documented in horse-race betting. On average, bettors tend to overvalue longshots and undervalue favorites. Researchers watch for this pattern in prediction markets too, though how strong it is in any specific market is still being studied.

Questions Without a Clear Answer

A crowd can only be right about a question that has a right answer. When the wording of a market is vague, traders may be betting on different interpretations, and the price blends those interpretations rather than one forecast. When a question has no precise answer, crowds can also drift toward arbitrary conclusions. This is why the rules for how a market settles matter so much. Our guide on how Polymarket markets resolve explains the role of the oracle and the resolution rules.

How Do We Know a Crowd Is Right? Calibration and the Brier Score

A single forecast cannot be judged right or wrong on its own. If a market said 70% and the event did not happen, that is not necessarily a mistake: 70% events are supposed to fail about three times in ten. To judge forecasters fairly, you look at many forecasts together.

The standard tool is the Brier score, introduced by Glenn Brier in 1950 in the journal Monthly Weather Review. For yes/no events, you take each forecast as a probability between 0 and 1, compare it to the outcome (1 if it happened, 0 if not), square the gap, and average across all forecasts:

Brier score = average of (forecast − outcome)²

It ranges from 0 (perfect) to 1 (as wrong as possible), so lower is better. A forecaster who always says 50% scores 0.25, which is a useful baseline.

A tiny worked example: you forecast 80% for an event that happens, giving (0.8 − 1)² = 0.04. You forecast 70% for an event that does not happen, giving (0.7 − 0)² = 0.49. Your average across both is (0.04 + 0.49) ÷ 2 = 0.265, slightly worse than always saying 50%.

The related idea is calibration. A well-calibrated forecaster or market is one where, among all events given a 70% chance, about 70% actually happen. The research on markets and superforecasters described above is largely about calibration: whether the probabilities match reality over many questions, not whether any single call was right.

When Crowds Are Wise vs. When They Are Not

Conditions that helpConditions that hurt
Many participants with different backgroundsA few traders, or everyone drawing on the same source
People form views independentlyPeople copy visible opinions or recent price moves
Local and specialized knowledge is rewardedInformation is concentrated or hidden
Deep liquidity, so single orders move prices littleThin markets where one order can swing the price
A clear question with an objective resolutionVague wording or no precise answer
Errors in different directions that cancel outA bias shared by the whole crowd

How to Read a Market Price Wisely

  • Treat the price as an estimate. A 62-cent contract is commonly read as a 62% chance, but that is the crowd's current best guess, not a promise.
  • Check volume and liquidity. A price backed by many traders and deep order books deserves more weight than one set by a few small trades.
  • Read the resolution rules. Make sure you know exactly what question the market is answering and how it will settle.
  • Watch for sudden spikes. A sharp move without news may reflect one large order or herding rather than new information.
  • Compare sources. Looking at polls, expert forecasts and other markets side by side helps you spot when one signal is out of line.

Key terms

  • Price discovery: the process by which trading pushes a price toward the level that balances buyers' and sellers' views. In prediction markets, it turns many private opinions into one public estimate.
  • Calibration: how well forecast probabilities match reality over many events. If 70% forecasts come true about 70% of the time, the forecaster is well calibrated.
  • Brier score: a number from 0 to 1 that measures the accuracy of probability forecasts on yes/no events. Lower is better; always guessing 50% scores 0.25.
  • Herding: when people follow others' choices instead of their own information, which reduces the independence a crowd needs to be wise.

Curious how a crowd sets a price in real time?

You can browse live Polymarket contracts and watch prices move as traders react to news, without trading anything. If you decide to go further, start with our beginner guide on how to trade on Polymarket.

Sources

  • Wikipedia, Wisdom of the crowd — https://en.wikipedia.org/wiki/Wisdom_of_the_crowd
  • Wikipedia, The Wisdom of Crowds (book) — https://en.wikipedia.org/wiki/The_Wisdom_of_Crowds
  • Hayek, The Use of Knowledge in Society (1945) — https://oll.libertyfund.org/titles/hayek-the-use-of-knowledge-in-society-1945
  • Rhode & Strumpf, Historical Presidential Betting Markets — https://doi.org/10.1257/0895330041371277
  • Wolfers & Zitzewitz, Prediction Markets — https://www.aeaweb.org/articles?id=10.1257%2F0895330041371321
  • Wikipedia, Iowa Electronic Markets — https://en.wikipedia.org/wiki/Iowa_Electronic_Markets
  • Good Judgment, About — https://goodjudgment.com/about/
  • Lorenz et al., How social influence can undermine the wisdom of crowd effect — https://pmc.ncbi.nlm.nih.gov/articles/PMC3107299
  • Rhode & Strumpf, Manipulating Political Stock Markets — https://users.wfu.edu/strumpks/papers/ManipIHT_June2008(KS).pdf
  • Slate on the Intrade 2012 trader — https://slate.com/news-and-politics/2013/09/2012-intrade-paper-suggests-a-single-intrade-trader-spent-millions-to-make-it-look-like-mitt-romney-could-win.html
  • History.com on tulip mania — https://www.history.com/articles/tulip-mania-financial-crash-holland
  • Wikipedia, Favorite-longshot bias — https://en.wikipedia.org/wiki/Favorite-longshot_bias
  • University of Virginia Library, A Brief on Brier Scores — https://library.virginia.edu/data/articles/a-brief-on-brier-scores